Simulation Validation Alternatives when Flight Test Data are Severely Limited

Abstract

Simulation validation involves gathering and analyzing evidence regarding a simulation’s suitability to represent a system’s behavior with sufficient accuracy for a specific purpose, such as verifying the system’s performance or evaluating a specific algorithm design. It is common practice on many programs—particularly those for which individual flight tests are very expensive—to release a simulation version for a single flight test. Without replication, it is difficult to make any statement regarding the statistical significance of variations between flight test data and simulation predictions. Ultimately, deciding simulation validity is an engineering judgment that is collaboratively rendered after subject matter experts compare data from various simulations and tests conducted under different conditions. But what can be said about the simulation, based upon data from a single flight test? This paper reviews analysis methods commonly employed in simulation validation. The benefits, required assumptions and other restrictions for goodness-of-fit tests are reviewed. A specific method is then proposed for use when the number of flight tests is severely limited. The proposed method begins with the selection of a dominant subset of simulation parameters, defined as flight-test matching parameters (FMPs), to focus simulation validation efforts. A three-phase analysis protocol is described for incremental simulation validation, during which FMP metrics are monitored. Typical patterns in FMP values are illustrated using results from validating an actual missile simulation.

Cite this paper

@article{hurst2008simulation,
  title = {{Simulation Validation Alternatives when Flight Test Data are Severely Limited}},
  author = {Hurst, Terril N. and Collins, Brett},
  journal = {AIAA Modeling and Simulation Technologies Conference},
  year = {2008},
  doi = {10.2514/6.2008-6357}
}

Made from the index record. Check names, capitalisation and page numbers against the paper. Need another style? AI for Research has over 10,000.

Related papers